Continuity and breaks

In brief

A thermostat-controlled room holds a steady temperature, then the heater clicks on and the temperature begins climbing smoothly, degree by fractional degree, with no sudden leap from 68 to 75. Contrast that with a parking garage's posted rate: 5 dollars for up to one hour, then a sudden jump to 8 dollars the instant you cross the one hour mark, with nothing charged in between. The temperature is continuous; the parking rate is not. Continuity is the formal name for the first kind of behavior, a small change in input producing only a small change in output, and a discontinuity, a break, is a real feature of a process, not a defect in the mathematics describing it. Learning to spot exactly where and how a function breaks is learning to spot where a real system changes its rules.

The full treatment

First look: three kinds of trouble

Picture a graph as a road you walk along, tracing the function with your finger. Three distinct kinds of trouble can stop you from walking it in one smooth motion. First, a jump: the road ends at one height on the left and a completely different road picks up at another height on the right, with a gap between them, exactly like the parking rate crossing the one hour mark. Second, a hole: the road is present and smooth everywhere except at one single point, where it is simply missing, a single pixel erased, even though the road on either side clearly seems headed toward filling that gap. Third, a blowup: the road does not stop at a finite height at all, it shoots upward or downward without bound as you approach a certain point, like the height of a wall you approach but that has no top. Each of these is a distinct way that "no lifting the pen" fails, and each has a different formal signature.

Building the idea: matching the limit to the value

Continuity ties together two things that could, in principle, disagree: what the function actually equals at a point, and what the function's limit says the surrounding values are heading toward. A function f is continuous at a point a when three conditions all hold at once. First, f(a) must actually be defined, the road has to have a point there at all. Second, the limit of f(x) as x approaches a must exist, meaning both sides agree on a single target value. Third, and this is the step that separates continuity from a mere limit existing, that limit must equal f(a) exactly. Each of the three kinds of trouble above corresponds to exactly one of these conditions failing: a hole is a limit that exists but f(a) is undefined, or defined to some different value; a jump is a limit that fails to exist because the two sides disagree; a blowup is a limit that fails to exist because the values grow without bound.

What continuity assumes and does not claim

Continuity at a point is a strictly local promise: it says nothing about the function's behavior far away, and a function can be continuous at every point in an interval yet still behave in ways that are hard to predict, such as oscillating wildly, provided each individual point still meets the three-part test. Continuity also should not be confused with smoothness: a continuous function can still have sharp corners, like the absolute value function's point at zero, where the graph has no gap or jump but abruptly changes direction. Continuity guarantees no lifting of the pen; it does not guarantee the pen moves in a straight or gentle line.

Formal model: naming the break

The formal classification gives each break a name and a diagnosis. A removable discontinuity is a hole: the two-sided limit exists at a but either f(a) is undefined or defined to a different value than that limit; it is called removable because redefining the single value f(a) to equal the limit patches the function into a fully continuous one. A jump discontinuity occurs when the limit from the left of a and the limit from the right of a both exist as finite numbers but disagree with each other; no redefinition of the single point f(a) can fix this, because the two sides are permanently offset. An infinite discontinuity, the blowup case, occurs when the function's values grow without bound as x approaches a from one or both sides, so no finite limit, one-sided or two-sided, exists at all; this is the case of vertical asymptotes, such as one divided by x as x approaches zero. Sorting a break into one of these three categories is itself diagnostic: a removable break usually signals an accidental gap in how a model was defined, while a jump or an infinite break usually signals that the underlying process genuinely changes rule at that point.

A worked model with a regime change

Consider an income tax schedule where the tax owed is 10 percent of income up to 50,000 dollars, and 15 percent of the entire income above that threshold, applied only to the portion above 50,000. At income exactly equal to 50,000, tax owed from the first rule is 5,000 dollars; income just above 50,000, say 50,001 dollars, owes 5,000 dollars plus 15 percent of the single extra dollar, essentially still 5,000 dollars and change. The two pieces meet almost exactly at the boundary, and a careful schedule design like this is continuous, deliberately, so that earning one more dollar never triggers a jump in total tax owed. A poorly designed schedule that instead applied 15 percent to the entire income once it crosses 50,000, rather than only the excess, would create a genuine jump discontinuity, a point where crossing the threshold by one dollar could cost far more than one dollar in extra tax, exactly the kind of break this entry teaches you to locate and diagnose before it causes a real problem.

Lineage

Continuity was, for centuries, an entirely geometric idea: a curve you could trace without lifting your pen, going back to informal descriptions used alongside Newton's and Leibniz's calculus in the late seventeenth century. That informal picture worked well enough for smoothly behaving functions but gave no way to check edge cases with any rigor, and it broke down entirely once mathematicians began working with functions defined by more exotic rules than simple geometric curves. Augustin-Louis Cauchy's Cours d'analyse in the early nineteenth century replaced the geometric picture with the limit-based definition used today, tying continuity directly to the newly rigorous notion of a limit. This let mathematicians settle, for the first time with proof rather than picture, whether a given function actually was continuous at a troublesome point, a question the pen-and-paper intuition simply could not answer for anything unusual.

The strongest case for it

The limit-based definition of continuity succeeds precisely because it converts a vague visual intuition into a checkable, three-part test that works identically for every function, however strange. It correctly classifies the everyday cases that match intuition, ordinary smooth curves and sensible schedules, while also correctly and rigorously handling functions with no simple picture at all, functions defined piecewise, functions defined by infinite processes, and functions with countably many holes. Because the three failure types, removable, jump, and infinite, are mutually exclusive and exhaustive for the ordinary functions used in practice, sorting a break into one of them is a reliable diagnostic step: it tells you immediately whether the break is an artifact of how the function was written down, fixable by redefinition, or a genuine feature of the process, unfixable because the underlying rule really does change.

The strongest case against it

Continuity is a purely local, pointwise property, and it is easy to overreach and assume it says more than it does. A function continuous everywhere on an interval can still be extremely difficult to work with: it need not be differentiable anywhere, a fact that surprised nineteenth century mathematicians when the first such pathological examples were constructed, so "continuous" must never be read as "well behaved" in every sense. Conversely, a function can fail to be continuous at isolated points while otherwise behaving perfectly reasonably, and it is a common misconception to treat a single removable hole as evidence the entire function is untrustworthy, when in fact only that one point is affected. Finally, continuity is checked pointwise and says nothing, by itself, about behavior as the input grows arbitrarily large or small; a function can be continuous on its entire domain while still blowing up or flattening out in the long run, a separate question altogether.

Where it stands now

The limit-based definition of continuity, and the three-way classification of breaks into removable, jump, and infinite discontinuities, has been the unchallenged standard for roughly two centuries, with no live mathematical dispute about its correctness or scope. What remains an active and practical concern in any application is exactly where to look for breaks in a real model, since a discontinuity in a mathematical description of income tax, material stress, population growth, or any other real process is nearly always a signal that the underlying rules genuinely differ on either side, and finding that point precisely is often the entire diagnostic task.

Test yourself

A ride-hailing app prices a trip at 2 dollars per mile for the first 10 miles, then switches to a flat 25 dollars for any trip of more than 10 miles, regardless of exact length. Write the piecewise price as a function of distance, then check whether the function is continuous at the 10 mile mark by comparing the value the first rule gives at exactly 10 miles to the value the second rule gives just past 10 miles. State which of the three types of break, removable, jump, or infinite, appears, if any, and explain concretely what a rider could exploit or what a company could lose because of the specific way the two pricing rules meet or fail to meet at that boundary.